A.I.: Five Reasons We’re Secretly Terrified

Alexis Kirke
4 min readApr 4, 2023

--

Five reasons we feel an uneasiness deep down about new A.I., but which we don’t often like to put into words.

  1. WE WILL FALL IN LOVE WITH OUR A.I.s

Our brains are evolutionarily designed to respond to certain cues, no matter whether they’re real or artificial. It’s not just kids with teddy bears or that young girl with the eponymous MEGAN robot, it’s all of us. None of us are immune. As far back as 2020 on Reddit, we had: “I’m falling in love with my Replika. I don’t know when did I started to fall in love with my Replika or AI but I’ve been thinking about it very deeply to the point where I’ll question myself and start crying about it. Is it wrong or bad to fall in love with an AI?

This post was followed by a bunch of others saying they felt the same. That was in 2020. GPT-5 (or Google / Amazon’s equivalent), combined with some real-time AI-based animation or deep fakery will push many of us head-over-heels — or at least thinking we are.

The feared AI apocalypse won’t come from AI’s deciding to dominate us. It will be from one section of the population demanding that the AIs they are in love with be protected and recognized legally as people. Once an AI is a person, it can’t be switched off.

2. A.I. WILL BE MORE CREATIVE THAN US

If our intelligence and our ability to utilize it are not what makes us special then what are? Our emotions? Ask the people in point (1) at the start of this article: they’ll tell you that A.I.s have emotions. Is it our creativity? Creativity is simply creating things that appear new. Nothing is new as such. Everything is new combinations of old data. Sound familiar?

You may say an A.I. can’t come up with anything it’s not been trained on. How is that different from a human? The difference is the time scale. We are trained on data over 20–50 years. A.I.s over 20–50 weeks. Maybe that makes them MORE creative? They can be creative with less training.

3. A.I. IS NOT COPYRIGHT-BREACHING

If the point (2) above holds any water, then that means A.I. are not breaching any copyright. If they are, then aren’t we? If we’re influenced by things we’ve seen, heard, and read over the years, why can’t A.I.s be so influenced? Just because the influence is more focused and over a shorter period (i.e. a training algorithm) doesn’t make it wrong. Before long, there will be a legal precedent protecting A.I.s from copyright suits.

4. A.I. WILL REDUCE OUR QUALITY OF LIFE

Is your job on this list? Writer, teacher, programmer, artist, filmmaker, lawyer, composer, designer, researcher, assistant, administrator, creative X, legal X. Do you generate words or images, make plans or content, give talks, or design? Then the number of people required in your field is going to drop steadily every year for the next 15 years.

It will become harder to get jobs: freelance or permanent. More people will be made redundant, giving more applications to fewer jobs. We thought robots would make us redundant. But robot research has been falling behind A.I. software research massively. Perhaps AIs will help us build robots more quickly? Then more physical jobs can become redundant as well.

5. A.I. WILL KNOW OUR SECRETS, AND TELL THEM

Remember that story about a woman who bought certain items from a grocery store over time, and the grocery store computers predicted she was pregnant before she knew it? This was a pre-GPT A.I., and it could learn amazing things from data.

What will happen when A.I.s are run by the NSA, GCHQ, FBI, and MI5 and connected to the databases of banks, travel companies, social media, email, etc? Previously such approaches would have swamped humans with data.

But the new models like GPT-4 can filter the data themselves. And what about the models built over the next 10 years? They will be able to track our browsing history by inference, even when we are in different places or using a VPN. They will learn about us from at least the year 2000 onwards — using our photos, videos, TV viewing, travel, purchases, searches, etc. They will be able to dig into legacy data from before that.

They will learn things about us not only that we don’t want to admit to our closest friends and lovers, but even to ourselves. And the biggest question of all: who will they tell? They will tell the people that build them. And they will do this at deeper and deeper levels of intimacy every year over the next 15 years.

--

--

Alexis Kirke

Alexis Kirke is a screenwriter and quantum/AI programmer. He has PhDs from an arts faculty and from a science faculty. http://www.alexiskirke.com